An Intelligent Multi-Classifier Framework for Cyberbullying Detection and User Moderation in Social Media

Abstract

Cyberbullying has emerged as a serious concern on social media platforms, negatively affecting individuals through abusive and offensive online interactions. The rapid growth of user-generated content makes manual monitoring ineffective, creating the need for intelligent automated detection systems. This paper presents a machine learning-based framework for detecting cyberbullying in textbased social media posts. The proposed system performs text preprocessing and TF-IDF feature extraction to convert raw textual data into meaningful features. Three machine learning algorithms, namely AdaBoost, Stochastic Gradient Descent (SGD), and Multinomial Naïve Bayes, are trained and evaluated to classify posts as offensive or non-offensive. Sentiment analysis is integrated to enhance the classification process by providing additional contextual information. Furthermore, the system monitors repeated offensive behavior and supports automated user moderation through an administrative interface. Experimental results demonstrate that the proposed framework provides reliable classification performance and contributes to creating a safer and more secure online communication environment.